Error in scale_index[[i]]: subscript out of bounds when we have layers with different datasets and facets with free scales

[smouksassi]

---

When having a composite ggplot with a data subset per layer setting scales = "free" gives an error.
What is fundamentally wrong with this approach ?
setting scales to fixed don't produce an error

require(ggplot2)
#> Loading required package: ggplot2
ggplot(mtcars, aes(mpg, disp)) + facet_grid(~cyl,scales="free") + 
  geom_point() +
  geom_point(data = subset(mtcars, cyl == 4)) +
  geom_line(data = subset(mtcars, cyl == 6)) +
  geom_text(data = subset(mtcars, cyl == 8), aes(label = gear))
#> Error in scale_index[[i]]: subscript out of bounds

ggplot(mtcars, aes(mpg, disp)) + facet_grid(~cyl,scales="fixed") + 
  geom_point() +
  geom_point(data = subset(mtcars, cyl == 4)) +
  geom_line(data = subset(mtcars, cyl == 6)) +
  geom_text(data = subset(mtcars, cyl == 8), aes(label = gear))

^{Created on 2019-01-24 by the reprex package (v0.2.1)}

[karawoo]

What version of ggplot2 are you using? I can't reproduce this error with 3.1.0.

[smouksassi]

other attached packages:
[1] ggplot2_3.1.0.9000 usethis_1.4.0 devtools_2.0.1

[karawoo]

Ah, looks like this was introduced in #3013. Having any one of the additional layers using subsetted data causes the error. The problem is here:

ggplot2/R/layout.R

Line 275 in 1f259f4

scale_index <- unname(split(seq_along(scale_id), scale_id))

Taking the following example:

library("ggplot2")
ggplot(mtcars, aes(mpg, disp)) + facet_grid(~cyl,scales="free") +
  geom_line(data = subset(mtcars, cyl == 6))
#> Error in scale_index[[i]]: subscript out of bounds

^{Created on 2019-01-24 by the reprex package (v0.2.1)}

Before the removal of plyr, scale_index looked like this:

scale_index
#> [[1]]
#> integer(0)

#> [[2]]
#> [1] 1 2 3 4 5 6 7

#> [[3]]
#> integer(0)

With the changes in #3013 only the non-empty list elements are included, which then causes the subsetting error in the next part of the function.

scale_index
#> [[1]]
#> [1] 1 2 3 4 5 6 7

[thomasp85]

Thanks for dissecting that @karawoo — I’ll make sure to fix it

[lock]

This old issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with reprex) and link to this issue. https://reprex.tidyverse.org/